In the recent drive for higher integration and operating speeds in LSI devices, the pattern rule is made drastically finer. The photolithography which is currently on widespread use in the art is approaching the essential limit of resolution determined by the wavelength of a light source. As the light source used in the lithography for resist pattern formation, g-line (436 nm) or i-line (365 nm) from a mercury lamp was widely used in 1980's. Reducing the wavelength of exposure light was believed effective as the means for further reducing the feature size. For the mass production process of 64 MB dynamic random access memories (DRAM, processing feature size 0.25 μm or less) in 1990's and later ones, the exposure light source of i-line (365 nm) was replaced by a KrF excimer laser having a shorter wavelength of 248 nm. However, for the fabrication of DRAM with a degree of integration of 256 MB and 1 GB or more requiring a finer patterning technology (processing feature size 0.2 μm or less), a shorter wavelength light source was required. Over a decade, photolithography using ArF excimer laser light (193 nm) has been under active investigation. It was expected at the initial that the ArF lithography would be applied to the fabrication of 180-nm node devices. However, the KrF excimer lithography survived to the mass-scale fabrication of 130-nm node devices. So, the full application of ArF lithography started from the 90-nm node. The ArF lithography combined with a lens having an increased numerical aperture (NA) of 0.9 is considered to comply with 65-nm node devices. For the next 45-nm node devices which required an advancement to reduce the wavelength of exposure light, the F2 lithography of 157 nm wavelength became a candidate. However, for the reasons that the projection lens uses a large amount of expensive CaF2 single crystal, the scanner thus becomes expensive, hard pellicles are introduced due to the extremely low durability of soft pellicles, the optical system must be accordingly altered, and the etch resistance of resist is low; the postponement of F2 lithography and the early introduction of ArF immersion lithography were advocated (see Proc. SPIE Vol. 4690, xxix, 2002).
In the ArF immersion lithography, the space between the projection lens and the wafer is filled with water. Since water has a refractive index of 1.44 at 193 nm, pattern formation is possible even using a lens having a numerical aperture (NA) of 1.0 or greater. Theoretically, it is possible to increase the NA to nearly 1.44. It was initially recognized that the resolution could be degraded and the focus be shifted by a variation of water's refractive index with a temperature change. The problem of refractive index variation could be solved by controlling the water temperature within a tolerance of 1/100° C. while it was recognized that the impact of heat from the resist film upon light exposure drew little concern. There was a likelihood that micro-bubbles in water could be transferred to the pattern. It was found that the risk of bubble generation is obviated by thorough deaeration of water and the risk of bubble generation from the resist film upon light exposure is substantially nil. At the initial phase in 1980's of the immersion lithography, a method of immersing an overall stage in water was proposed. Later proposed was a partial-fill method of using a water feed/drain nozzle for introducing water only between the projection lens and the wafer so as to comply with the operation of a high-speed scanner. In principle, the immersion technique using water enabled lens design to a NA of 1 or greater. In optical systems based on traditional refractive index materials, this leads to giant lenses, which would deform by their own weight. For the design of more compact lenses, a catadioptric system was proposed, accelerating the lens design to a NA of 1.0 or greater. A combination of a lens having NA of 1.2 or greater with strong resolution enhancement technology suggests a way to the 45-nm node (see Proc. SPIE Vol. 5040, 724, 2003). Efforts have also been made to develop lenses of NA 1.35.
One candidate for the 32-nm node lithography is lithography using extreme ultraviolet (EUV) radiation with wavelength 13.5 nm. The EUV lithography has many accumulative problems to be overcome, including increased laser output, increased sensitivity, increased resolution and minimized line-width roughness (LWR) of resist coating, defect-free MoSi laminate mask, reduced aberration of reflection mirror, and the like.
The water immersion lithography using a NA 1.35 lens achieves an ultimate resolution of 40 to 38 nm as the half-pitch of a line pattern at the maximum NA, but cannot reach 32 nm. Efforts have been made to develop higher refractive index materials in order to further increase NA. It is the minimum refractive index among projection lens, liquid, and resist film that determines the NA limit of lenses. In the case of water immersion, the refractive index of water is the lowest in comparison with the projection lens (refractive index 1.5 for synthetic quartz) and the resist film (refractive index 1.7 for prior art methacrylate-based film). Thus the NA of projection lens is determined by the refractive index of water. Recent efforts succeeded in developing a highly transparent liquid having a refractive index of 1.65. In this situation, the refractive index of projection lens made of synthetic quartz is the lowest, suggesting a need to develop a projection lens material with a higher refractive index. LuAG (lutetium aluminum garnet Lu3Al5O12) having a refractive index of at least 2 is the most promising material, but has the problems of birefringence and noticeable absorption. Even if a projection lens material with a refractive index of 1.8 or greater is developed, the liquid with a refractive index of 1.65 limits the NA to 1.55 at most, failing in resolution of 32 nm despite successful resolution of 35 nm. For resolution of 32 nm, a liquid with a refractive index of 1.8 or greater and resist and protective films with a refractive index of 1.8 or greater are necessary. Among the materials with a refractive index of 1.8 or greater, the high refractive index liquid seems least available. Such a liquid material has not been discovered because a tradeoff between absorption and refractive index is recognized in the art. In the case of alkane compounds, bridged cyclic compounds are preferred to linear ones in order to increase the refractive index, but the cyclic compounds undesirably have too high a viscosity to follow high-speed scanning on the exposure tool stage. Since hafnium oxide particles have high transparency and a refractive index in excess of 2 at 193 nm, it is under study to form a high refractive index liquid by dispersing the particles in water or alkane solvents. However, to increase the refractive index up to 1.8, hafnium oxide must be dispersed in water in an amount of at least 30 wt %. The resulting mixture has a very high viscosity which is incompatible with high-speed scanning. If a liquid with a refractive index of 1.8 is developed, then the component having the lowest refractive index is the resist film, suggesting a need to increase the refractive index of a resist film to 1.8 or higher.
The process that now draws attention under the above-discussed circumstances is a double patterning process involving a first set of exposure and development to form a first pattern and a second set of exposure and development to form a pattern between the first pattern features. See Proc. SPIE Vol. 5754, 1508 (2005). A number of double patterning processes are proposed. One exemplary process involves a first set of exposure and development to form a photoresist pattern having lines and spaces at intervals of 1:3, processing the underlying layer of hard mask by dry etching, applying another layer of hard mask thereon, a second set of exposure and development of a photoresist film to form a line pattern in the spaces of the first exposure, and processing the hard mask by dry etching, thereby forming a line-and-space pattern at a half pitch of the first pattern. An alternative process involves a first set of exposure and development to form a photoresist pattern having spaces and lines at intervals of 1:3, processing the underlying layer of hard mask by dry etching, applying a photoresist layer thereon, a second set of exposure and development to form a second space pattern on the remaining hard mask portion, and processing the hard mask by dry etching. In either process, the hard mask is processed by two dry etchings.
While the former process requires two applications of hard mask, the latter process uses only one layer of hard mask, but requires to form a trench pattern which is difficult to resolve as compared with the line pattern. The latter process includes the use of a negative resist material in forming the trench pattern. This allows for use of high contrast light as in the formation of lines as a positive pattern. However, since the negative resist material has a lower dissolution contrast than the positive resist material, a comparison of the formation of lines from the positive resist material with the formation of a trench pattern of the same size from the negative resist material reveals that the resolution achieved with the negative resist material is lower. After a wide trench pattern is formed from the positive resist material by the latter process, there may be applied a thermal flow method of heating the substrate for shrinkage of the trench pattern, or a RELACS method of coating a water-soluble film on the trench pattern as developed and heating to induce crosslinking at the resist film surface for achieving shrinkage of the trench pattern. These have the drawbacks that the proximity bias is degraded and the process is further complicated, leading to reduced throughputs.
Both the former and latter processes require two etchings for substrate processing, leaving the issues of a reduced throughput and deformation and misregistration of the pattern by two etchings.
One method that proceeds with a single etching is by using a negative resist material in a first exposure and a positive resist material in a second exposure. Another method is by using a positive resist material in a first exposure and a negative resist material in a higher alcohol of 4 or more carbon atoms, in which the positive resist material is not dissolvable, in a second exposure. However, these methods using negative resist materials with low resolution entail degradation of resolution.
A method which does not involve PEB and development between first and second exposures is the simplest method with high throughput. This method involves first exposure, replacement by a mask having a shifted pattern drawn, second exposure, PEB, development and dry etching. However, the optical energy of second exposure offsets the optical energy of first exposure so that the contrast becomes zero, leading to a failure of pattern formation. If an acid generator capable of two photon absorption or a contrast enhancement layer (CEL) is used to provide nonlinear acid generation, then the energy offset is relatively reduced even when second exposure is performed at a half-pitch shifted position. Thus a pattern having a half pitch corresponding to the shift can be formed, though at a low contrast. See Jpn. J. Appl. Phy. Vol. 33 (1994) p 6874-6877, Part 1, No. 12B, December 1994.
The critical issue associated with double patterning is an overlay accuracy between first and second patterns. Since the magnitude of misregistration is reflected by a variation of line size, an attempt to form 32-nm lines at an accuracy of 10%, for example, simply requires an overlay accuracy within 3.2 nm. The advanced ArF immersion lithography scanner has an overlay accuracy of the order of 8 to 6 nm for every wafer on a common exposure tool. The term “every wafer” means that the exposure tool carries out alignment relative to a resist alignment pattern which has been formed by exposure and development. Herein a significant improvement in accuracy is necessary. If first and second exposures are carried out without demounting the wafer from the chuck, the positional shift associated with chuck remounting is cancelled and the alignment accuracy is improved to the order of 5 to 4 nm. In the case of double patterning, the process of carrying out plural exposures without demounting the wafer from the chuck becomes an exposure process which can be implemented on account of the improvement in alignment accuracy. In the exposure process intended to reduce the pitch to half by utilizing a nonlinear energy distribution such as the two-photon absorption resist, continuous exposure is carried out while shifting the exposure position by ¼ of the pitch and without demounting the wafer from the chuck. The nonlinear resist or CEL which is sensitive to radiation of 193 nm wavelength has not been reported. If such a resist were developed, the double exposure process with the minimum alignment error would become practical.
In addition to the double patterning technique, the technology for forming a fine space pattern or hole pattern includes use of negative resist material, thermal flow, and RELACS as mentioned above. The negative resist material suffers from the problems that the resist material itself has a low resolution and bridges form in a fine hole pattern because the negative resist material relies on the crosslinking system. The thermal flow and RELACS methods suffer from a likelihood of variation during dimensional shrinkage by heat.
FIG. 1 illustrates a process for forming a hole pattern using a positive photoresist material. In FIG. 4A, a photoresist material is coated onto a processable substrate 101 on a substrate 100 to form a photoresist film 102. In FIG. 4B, the photoresist film 102 is exposed to light through a photomask having the desired pattern and developed to form a photoresist pattern 102a. In FIG. 4C, the processable substrate 101 is etched while using the photoresist pattern 102a as a mask.
The method of forming a negative pattern by reversal of a positive pattern is well known from the past. For example, JP-A 2-154266 and JP-A 6-27654 disclose naphthoquinone resists capable of pattern reversal. JP-A 64-7525 describes exposure of selected portions of a film to focused ion beam (FIB) for curing and flood exposure whereby the cured portions are left behind. JP-A 1-191423 and JP-A 1-92741 describe exposure of a photosensitive agent of naphthoquinone diazide to form an indene carboxylic acid, heat treatment in the presence of a base into an indene which is alkali insoluble, and flood exposure to effect positive/negative reversal. FIG. 2 illustrates this positive/negative reversal method. In FIG. 2A, a photoresist material is coated onto a processable substrate 101 on a substrate 100 to form a photoresist film 102. In FIG. 2B, the photoresist film 102 is exposed to light through a photomask having the desired pattern and heated. In FIG. 2C, the photoresist film 102 is subjected to flood exposure. FIG. 2D illustrates pattern reversal by development to form a reversed pattern film 103. In FIG. 2E, the processable substrate 101 is etched while using the reversed pattern film 103 as a mask.
As to the positive/negative reversal method including exchange of developers, attempts were made to form negative patterns by development in an organic solvent of hydroxystyrene partially protected with tert-butoxycarbonyl (t-BOC) groups, and by development with super-critical carbon dioxide.
As to the positive/negative reversal method utilizing silicon-containing materials, it is proposed to form a fine hole pattern by covering a space portion of a positive resist pattern with a silicon-containing film, effecting oxygen gas etching for etching away the positive pattern portion, thus achieving positive/negative reversal to leave a silicon-containing film pattern. See JP-A 2001-92154 and JP-A 2005-43420. FIG. 3 illustrates this positive/negative reversal method. In FIG. 3A, a photoresist material is coated onto an underlayer film 104 on a processable substrate 101 on a substrate 100 to form a photoresist film 102. In FIG. 3B, the photoresist film 102 is exposed to light through a photomask having the desired pattern and developed to form a photoresist pattern 102a. In FIG. 3C, the photoresist pattern 102a is crosslinked. In FIG. 3D, a SOG film 105 is formed on the underlayer film 104 and so as to cover the crosslinked photoresist pattern 102a. FIG. 3E illustrates light etching with CMP or CF gas until the crosslinked photoresist pattern 102a is exposed. FIG. 3F illustrates pattern reversal by oxygen or hydrogen gas etching. In FIG. 3G, the processable substrate 101 is etched while using the patterned SOG film 105 as a mask.
As compared with the line pattern, the hole pattern is difficult to reduce the feature size. In order for the prior art method to form fine holes, an attempt is made to form fine holes by under-exposure of a positive resist film combined with a hole pattern mask, resulting in the exposure margin being extremely narrowed. It is then proposed to form holes of greater size, followed by thermal flow or RELACS method to shrink the holes as developed. However, there is a problem that control accuracy becomes lower as the pattern size after development and the size after shrinkage are greater and the quantity of shrinkage is greater. It is then proposed in Proc. SPIE, Vol. 5377, 255 (2004) that a pattern of X direction lines is formed using a positive resist film and dipole illumination, the resist pattern is cured, another resist material is coated thereon again, and a pattern of Y direction lines is formed in the other resist using dipole illumination, leaving a grid line pattern, spaces of which provide a hole pattern. Although a hole pattern can be formed at a wide margin by combining X and Y lines and using dipole illumination featuring a high contrast, it is difficult to etch vertically staged line patterns at a high dimensional accuracy. It is proposed in IEEE IEDM Tech. Digest 61 (1996) to form a hole pattern by exposure of a negative resist film through a Levenson phase shift mask of X direction lines combined with a Levenson phase shift mask of Y direction lines. Since the maximum resolution of ultrafine holes is determined by the bridge margin, the crosslinking negative resist film has the drawback that the threshold size is large as compared with the positive resist film.
As compared with white spots, a pattern of black spots can be formed to a fine size. As the hole size is reduced using a highly transparent halftone or chromeless phase shift mask, white spots are reversed into black spots, after which very small black spots with a high contrast are formed. It is reported in Proc. SPIE Vol. 4000, 266 (2000) to form dense fine holes by combining this concept with a negative resist film. A combination of a halftone phase shift mask having a high transmittance of 20% with a negative resist film provides a mask error enhancement factor (MEEF) of 0 (see Proc. SPIE Vol. 5040, 1258 (2003)).
When the super-resolution technology is applied to repeating dense patterns, the pattern density bias between dense and isolated patterns, known as proximity bias, becomes a problem. As the super-resolution technology used becomes stronger, the resolution of a dense pattern is more improved, but the proximity bias is exaggerated. In particular, an increase of proximity bias in a hole pattern poses a serious problem. One common approach taken to suppress the proximity bias is by biasing the size of a mask pattern. Since the proximity bias varies with properties of a photoresist material, specifically dissolution contrast and acid diffusion, the proximity bias of a mask varies with the type of photoresist material. For a particular type of photoresist material, a mask having a different proximity bias must be used. This adds to the burden of mask manufacturing. Then the pack and unpack (PAU) method is proposed in Proc. SPIE Vol. 5753, 171 (2005), which involves strong super-resolution illumination of a first positive resist to resolve a dense hole pattern, coating on the first positive resist pattern a negative resist film material in alcohol solvent which does not dissolve the first positive resist pattern, exposure and development of an unnecessary hole portion to close the corresponding holes, thereby forming both a dense pattern and an isolated pattern. One problem of the PAU method is misalignment between first and second exposures, as the authors point out in the report. The hole pattern which is not closed by the second development experiences two developments and thus undergoes a size change, which is another problem.